CNN-Based Intelligent Safety Surveillance in Green IoT Applications

Safety surveillance is considered one of the most important factors in many constructing industries for green internet of things(IoT)applications.However,traditional safety monitoring methods require a lot of labor source.In this paper,we propose intelligent safety surveillance(ISS)method using a convolutional neural network(CNN),which is an autosupervised method to detect workers whether or not wearing helmets.First,to train the CNN-based ISS model,the labeled datasets mainly come from two aspects:1)our labeled datasets with the full labeled on both helmet and pedestrian;2)public labeled datasets with the parts labeled either on the helmet or pedestrian.To fully take advantage of all datasets,we redesign CNN structure of network and loss functions based on YOLOv3.Then,we test our proposed ISS method based on the specific detection evaluation metrics.Finally,experimental results are given to show that our proposed ISS method enables the model to fully learn the labeled information from all datasets.When the threshold of intersection over union(IoU)between the predicted box and ground truth is set to 0.5,the average precision of pedestrians and helmets can reach 0.864 and 0.891,respectively.     

Leveraging data fusion to improve barrier coverage in wireless sensor networks

Intruder detection and border surveillance are amongst the most promising applications of wireless sensor networks. Barrier coverage formulates these problems as constructing barriers in a long-thin region to detect intruders that cross the region. Existing studies on this topic are not only based on simplistic binary sensing model but also neglect the collaboration employed in many systems. In this paper, we propose a solution which exploits the collaboration of sensors to improve the performance of barrier coverage under probabilistic sensing model. First, the network width requirement, the sensor density and the number of barriers are derived under data fusion model when sensors are randomly distributed. Then, we present an efficient algorithm to construct barriers with a small number of sensors. The theoretical comparison shows that our solution can greatly improve barrier coverage via collaboration of sensors. We also conduct extensive simulations to demonstrate the effectiveness of our solution.     

A Nonparametric Approach to Foreground Detection in Dynamic Backgrounds

Foreground detection is a fundamental step in visual surveillance.However,accurate foreground detection is still a challenging task especially in dynamic backgrounds.In this paper,we present a nonparametric approach to foreground detection in dynamic backgrounds.It uses a history of recently pixel values to estimate background model.Besides,the adaptive threshold and spatial coherence are introduced to enhance robustness against false detections.Experimental results indicate that our approach achieves better performance in dynamic backgrounds compared with several approaches.     

A Digital Evidence Fusion Method in Network Forensics Systems with Dempster-Shafer Theory

Network intrusion forensics is an important extension to present security infrastructure,and is becoming the focus of forensics research field.However,comparison with sophisticated multi-stage attacks and volume of sensor data,current practices in network forensic analysis are to manually examine,an error prone,labor-intensive and time consuming process.To solve these problems,in this paper we propose a digital evidence fusion method for network forensics with Dempster-Shafer theory that can detect efficiently computer crime in networked environments,and fuse digital evidence from different sources such as hosts and sub-networks automatically.In the end,we evaluate the method on well-known KDD Cup1999 dataset.The results prove our method is very effective for real-time network forensics,and can provide comprehensible messages for a forensic investigators.     

Intelligent Passive Detection of Aerial Target in Space-Air-Ground Integrated Networks

Passive detection of moving target is an important part of intelligent surveillance. Satellite has the potential to play a key role in many applications of space-air-ground integrated networks(SAGIN). In this paper, we propose a novel intelligent passive detection method for aerial target based on reservoir computing networks. Specifically, delayed feedback networks are utilized to refine the direct signals from the satellite in the reference channels. In addition, the satellite direct wave interference in the monitoring channels adopts adaptive interference suppression using the minimum mean square error filter. Furthermore, we employ decoupling echo state networks to predict the clutter interference in the monitoring channels and construct the detection statistics accordingly. Finally, a multilayer perceptron is adopted to detect the echo signal after interference suppression. Extensive simulations is conducted to evaluate the performance of our proposed method. Results show that the detection probability is almost 100% when the signal-to-interference ratio of echo signal is-36 dB, which demonstrates that our proposed method achieves efficient passive detection for aerial targets in typical SAGIN scenarios.     

Visual focus of attention estimation based on improved hybrid incremental dynamic Bayesian network

In this paper, a visual focus of attention(VFOA) detection method based on the improved hybrid incremental dynamic Bayesian network(IHIDBN) constructed with the fusion of head, gaze and prediction sub-models is proposed aiming at solving the problem of the complexity and uncertainty in dynamic scenes. Firstly, gaze detection sub-model is improved based on the traditional human eye model to enhance the recognition rate and robustness for different subjects which are detected. Secondly, the related sub-models are described, and conditional probability is used to establish regression models respectively. Also an incremental learning method is used to dynamically update the parameters to improve adaptability of this model. The method has been evaluated on two public datasets and daily exper iments. The results show that the method proposed in this paper can effectively estimate VFOA from user, and it is robust to the free deflection of the head and distance change.     

NEAR OPTIMAL CLUSTER-HEAD SELECTION FOR WIRELESS SENSOR NETWORKS

Clustering in wireless sensor networks is an effective way to save energy and reuse band- width. To our best knowledge, most of the clustering protocols proposed in literature are of a dynamic type, where cluster heads are selected in each period, followed by cluster formation. In this paper, a new static type clustering method called Hausdorff clustering, which is based on the location of sensor nodes as well as communication efficiency and network connectivity, is proposed. The cluster head, however, is rotated within the cluster by a fuzzy logic algorithm that optimizes the network lifetime. Simulation results show that this approach can significantly increase the lifetime of the sensor network.     

叠加电桥检测的加速度传感器的设计

This paper describes the design,simulation,processing and test result of a high sensitivity accelerometer based on the piezoresistive effect which uses an overlay bridge detection method.The structure of this accelerometer is supersymmetric "mass-beams".This accelerometer has 8 beams,where two varistors are put in the two ends.Four varistors compose a Wheatstone bridge and the output voltages of the 4 Wheatstone bridges have been superimposed as the final output voltage.The sensitivity of the accelerometer can be improved effectively by these clever methods. A simplified mathematical model has been created to analyze the mechanical properties of the sensor,then the finite element modeling and simulation have been used to verify the feasibility of the accelerometer.The results show that the sensitivity of the accelerometer is 1.1381 m V/g,which is about four times larger than that of the single bridge accelerometers and series bridge sensor.The bandwidth is 0-1000 Hz which is equal to that of the single bridge accelerometers and the series bridge sensor.The comparison reveals that the new overlay detection bridge method can improve the sensitivity of the sensor in the same bandwidth.Meanwhile,this method provides an effective method to improve the sensitivity of piezoresistive sensors.     

Neural network based method for background modeling and detecting moving objects

This paper proposes a novel method, primarily based on the fuzzy adaptive resonance theory(ART) neural network with forgetting procedure, for moving object detection and background modeling in natural scenes. With the ability, inheriting from the ART neural network, of extracting patterns from arbitrary sequences, the background model based on the proposed method can learn new scenes quickly and accurately. To guarantee that a long-life model can derived from the proposed mothed, a forgetting procedure is employed to find the neuron that needs to be discarded and reconstructed, and the finding procedure is based on a neural network which can find the extreme value quickly. The results of a suite of quantitative and qualitative experiments conducted verify that for processes of modeling background and detecting moving objects our method is more effective than five other proven methods with which it is compared.     

Double-branch forgery image detection based on multi-scale feature fusion

Most of existing methods exhibit poor performance in detecting forged images due to the small size of tampered areas and the limited pixel difference between untampered and tampered regions. To alleviate the above problem, a double-branch tampered image detection based on multi-scale features is proposed. Firstly, we introduce a fusion module based on attention mechanism in the first branch to enhance the network’s sensitivity towards tampered regions. Secondly, we construct a second branch spec...     

Visual node module: An open‐source extension to the ns‐3 network simulator

A wireless visual sensor network is a collective network of directional and battery‐operated sensor nodes equipped with cameras. The field of view of these nodes depends on the camera opening angle, its direction, and its depth of view. Therefore, coverage and object detection in this type of networks are more challenging compared with the traditional wireless sensor networks. Thus, many researchers propose algorithms and solutions in this field that need tests and simulations. In this paper, we focus on network simulator 3 (ns‐3), which is an open‐source and discrete‐event tool suitable for wireless network simulation targeted primarily for research and educational use. The lack of models that can simulate visual sensor nodes in this simulator motivated us to design and develop a new visual node module as an extension of the ns‐3 core libraries and also to adapt the NetAnim tool to present these nodes graphically. This module will help researchers to simulate, test, and visualize their solutions in wireless visual sensor networks field. In this paper, we present the design and implementation of the proposed module. Furthermore, we show how it can be used in ns‐3 to simulate different scenarios of object detection and visualize the results in NetAnim tool.     

基于深度卷积神经网络的红外图像行人检测

红外图像中的行人检测一直是计算机视觉领域的研究热点与难点。针对传统的红外行人检测方法需要人工设计目标表达特征的弊端,本文从深度学习的角度出发,提出一种可以自动构建目标表达特征的红外行人检测卷积神经网络。在对卷积神经网络的实现原理进行分析的基础上,设计了红外行人检测卷积神经网络的初始结构,然后通过实验对初始结构进行调整,得到最终的检测神经网络。对实拍红外人体数据库进行行人检测的实验结果表明,该方法在保持低虚警率的同时可以对红外图像中的行人进行稳健检测,优于传统方法。     

基于惯性传感器的动态步长算法研究

行人步态参数的精确估计是行人自主导航系统和行人健康监测的关键技术之一。针对当前行人自主导航系统中步长估算算法精度低和弱适应性的问题,提出了一种计算行人动态步长算法。首先对行人的步态特征进行分解,利用改进的零速检测确定行人运动状态,采用卡尔曼滤波技术降低惯性传感器中累积误差的影响,再对进行滤波和坐标转换后的加速度进行双重积分,最终得到行人脚尖的运动轨迹。通过采用MTI-700惯性模块设计实验并进行实验验证。结果表明,该文提出的步长算法计算的步长与行人实际步长的误差低于3.0%。与现有的行人动态步长算法相比,该算法首次计算出行人脚尖的运动轨迹,精度较高且适应强,在行人自主导航及行人健康监测领域具有较大的应用价值。     

基于YOLO的轻量红外图像行人检测方法

红外图像中行人的快速检测一直是计算机视觉领域的热点和难点。针对红外图像行人目标检测算法检测速度和检测精度难以平衡,算法模型体积较大,在中低性能设备中难以部署和实时运行的问题,提出了一种基于YOLO算法的轻量红外图像行人检测方法。在分析了MobileNet v3等轻量网络在YOLO v3算法上的性能和特点之后,该方法提出了引入注意力机制的轻量特征提取网络(CSPmini a)、特征融合模块和解耦检测端分类回归结构三种改进措施,在满足网络模型轻量的情况下保证了一定的检测精度。实验表明,该方法有效的实现了红外图像行人目标检测的准确性和快速性。     

面向目标跟踪的传感器网络布局优化及保护策略

本文以目标跟踪为应用背景，改进了已有传感器感知模型和虚拟力方法，提出了一种新的传感器网络布局优化策略，该策略首先计算传感器与目标、热点区域、障碍物和其他传感器之间的虚拟力，为各传感器寻找受力平衡点，并将其作为该传感器的新位置，从而优化网络布局优化．实验证明，该策略可有效改善传感器网络覆盖率和目标探测概率．同时，本文根据各传感器获取的信息量，提出了涉及目标的传感器网络节点重要性排序算法，以及根据节点重要性进行传感器网络保护的策略。     

Convolutional neural networks for multispectral pedestrian detection

In this paper, we present a pedestrian detection method by leveraging multispectral images which consist of color and thermal image information. Our method is based on the observation that a multispectral image enables us to overcome inherent limitations for pedestrian detection under challenging situations, e.g., insufficient illumination, small size pedestrian instances and occlusion. In order to detect pedestrian under such conditions, we apply deep convolutional neural networks (CNNs) for effectively combining color and thermal image information in multispectral images. We present a novel multispectral network that is built from the region-based fully convolutional networks (R-FCN) network model. A network-in-network (NIN) is employed to fuse these information across different modalities. Experimental results on KAIST benchmark demonstrate that our method surpasses the baseline method R-FCN and other proposed architectures.     

Optimal barrier coverage for critical area surveillance using wireless sensor networks

Intrusion detection using barrier coverage is one of many applications existed in wireless sensor networks. The main purpose of using barrier coverage is to monitor the borders of a specific area against the intruders that are trying to penetrate this critical area by ensuring the total coverage with a low cost and extending the lifetime of the network, many solutions have been proposed in the literature in order to solve the coverage problem in wireless sensor networks, which became a vital field of research. In this paper, we present a new technique based on geometric mathematical models, in order to guarantee the total coverage of our deployed barriers with a minimum possible number of sensors. The idea is to calculate the number of sensors adequate to cover a barrier before deployment. We then formulate the problem to minimize the number of sensors to be deployed to properly cover a barrier; the calculation makes it possible to solve this problem in polynomial using our own heuristic. Additionally, we propose a new mechanism for ensuring a fault‐tolerant network by detecting the faulty sensors and select other suited sensors to close the existing gaps inside the barriers and detecting the sensors whose energy is low before the failure. The obtained simulation results prove the effectiveness of the proposed algorithms.     

室内外行人导航系统研究

采用微机电系统(MEMS)惯性传感器、MEMS磁传感器及小型全球定位系统(GPS)接收机为室内外行人导航数据源,基于Cortex-M4为内核,搭建了室内外行人导航系统硬件平台。重点研究了多传感器导航系统的结构、多源信息融合方法、多条件零速检测方法及零速修正等理论方法。并通过试验,采集实测数据进行分析、验证行人导航系统设计的性能。结果表明,在GPS信号良好情况下,定位误差在2.5m以内;无GPS信号期间,路线长度为110m时,定位误差在总路经的5%内。     

Illumination-aware window transformer for RGBT modality fusion

Combination of RGB and thermal sensors has been proven to be useful for many vision applications. However, how to effectively fuse the information of two modalities remains a challenging problem. In this paper, we propose an Illumination-Aware Window Transformer (IAWT) fusion module to handle the RGB and thermal multi-modality fusion. Specifically, the IAWT fusion module adopts a window-based multi-modality attention combined with additional estimated illumination information. The window-based multi-modality attention infers dependency cross modalities within a local window, thus implicitly alleviate the problem caused by weakly spatial misalignment of the RGB and thermal image pairs within specific dataset. The introduction of estimated illumination feature enables the fusion module to adaptively merge the two modalities according to illumination conditions so as to make full use of the complementary characteristics of RGB and thermal images under different environments. Besides, our proposed fusion module is task-agnostic and data-specific, which means it can be used for different tasks with RGBT inputs. To evaluate the advances of the proposed fusion method, we embed the IAWT fusion module into different networks and conduct the experiments on various RGBT tasks, including pedestrian detection, semantic segmentation and crowd counting. Extensive results demonstrate the superior performance of our method.     

Vision-aware target recognition toward autonomous robot by Kinect sensors

Target recognition is a key module in modern human–computer interaction (HCI) and computer vision systems It is pervasively used in many domains like autonomous vehicles and robot, remote operation, and video surveillance. However, due to the complicated environment and object occlusion, target recognition is still a challenging task. In this paper, we propose a novel target recognition algorithm toward autonomous robot by leveraging the Kinect sensors. More specifically, we utilize the Kinect sensors to capture scenario image in real-time. Subsequently, we present an improved HSV-based image segmentation algorithm to decompose the captured image, where morphological operation is employed for foreground target extraction. Afterward, we leverage Spatial Pyramid (SP)-based scheme for visual feature extraction. Then, we adopt a new distance metric for target matching. Comprehensive experimental results have shown the effectiveness of our proposed method.